Portfolio Showcase

Work & Projects

Loading projects from the backend.

System Design & Architecture

AI Engineering Pipelines

Interactive blueprints of production-grade AI architectures designed and implemented by Hasmukh Mer.

Gatex AI & Knowledge Bases

Production architecture for high-precision document search and factual answering

Latency

~320ms

Accuracy

98.7% Precision

Efficiency

45% Token Savings

STEP 0115ms

Query & Intent Routing

Semantic Router / BERT

STEP 0245ms

Hybrid Retrieval

Qdrant Vector DB + BM25

STEP 0360ms

Cross-Encoder Re-Ranking

Cohere / BGE-Reranker

STEP 04200ms

LLM Generation & Guardrails

DeepSeek / GPT-4o / FastAPI

Step 01 Architecture Deep-Dive

Tech: Semantic Router / BERT · 15ms

Classifies user intent, checks query guardrails, and extracts semantic search keywords.

Hypothetical Document Embeddings (HyDE) expansion

Prompt injection filtering & sanitization

Intent classification into knowledge vs. code vs. general